Meta Passthrough Stack

Passthrough is not one thing. A compositor passthrough layer can make the real room visible behind virtual content without giving the app raw camera pixels. The Passthrough Camera API is a different surface: it exposes forward-facing RGB camera frames for computer-vision and machine-learning use cases under device, OS, permission, and policy constraints. Environment depth is separate again, and is mainly useful for occlusion, raycasting, and grounding virtual content in the room.

That separation matters for implementation and for public claims. The local notes use Meta and Khronos documentation, sample projects, view-synthesis papers, and patents as a source map. Patents and research systems are treated as design signals, not as evidence that a shipped headset exposes a hidden capability.

The practical thread is Rust and Unity experimentation around what can be built with the exposed stack. The public page keeps that at the level of documentation and source anchors; private implementation notes stay in the working repos.

Mixed-reality writing often collapses passthrough visibility into camera access. This project keeps those apart. Seeing the room through a headset, compositing virtual content over it, receiving camera frames, receiving depth, and inferring scene structure are related but different affordances.

The public usefulness of the archive is therefore corrective: it blocks vague claims before they turn into design assumptions. It also gives experimental work a cleaner checklist for what must be tested on-device rather than inferred from marketing language, patents, or old forum answers.